This is a classifier which can generate symbolic link of certain person's photo files in specified folder, to make the selection of certain person's photo easier. It's based on Adam Geitgey's Face Recognition project. The purpose of this project is to make selection of certian person's photo easier, giving you have lots of photos and mutliple directories hirarchies. Maybe you want to choose some photoes to print out, but some product, such as Google Photo, needs to upload photo, and sometimes it just not work very quickly, or even cannot work in some location. Looks KNN is better, according to the test result. KNN is more exact in case of multiple person photo and less time consuming. But currently the code just support to run on single CPU.
Classify photos according to the persons who we pay attention, the result is each concerned person has its own folder, its solo photo will have symbol link there, also a mutliple person photo foler, where also contains symbol link to the real photo.
- Same as Face Recognition
Download the code of this project, run it with arguments:
python3 path_to_classifier.py --cpus -1 path_to_symbollink_folder path_to_single_target_persons_photoes_folder path_to_photoes_to_classifyThe tolerance argument defaults to 0.45, which sounds reasonable to Asian person, you can use --tolerance VALUE to specify one in your command line. In project Face Recognition, default tolerance is 0.6.
When the classification is done, you can select certain photoes' symbol link files to copy to some folderA.
After that you can run cp -rL folderA folderB to copy these selected symbol link files' photoes to some folderB, then you can use them to publish your album or do whatever you want.
Many thanks to Adam Geitgey (@ageitgey) for creating the great face recognition API. If you love machine learning, pay attention to his blog post